Performance Tweaking about lz-string HOT 15 CLOSED

If you've patched the lib, can you make a jsperf out of it (against the current release version) ?

from lz-string.

Haven't implemented those changes currently. I'll see what I can do in my spare time when I'm off for Christmas, though.

from lz-string.

ok, thanks. Let us know.

from lz-string.

Just saw another slight tweak that I'll test out when I get to the rest of this. "context_w = String(context_c);" creates a string object instead of a string literal, which is going to be slower to work with. I'll try changing it to "context_w = '' + context_c;" to get around that.

from lz-string.

Alright, made a jsperf for it. The results seem about even, though I have IE11, so I'm not sure if this is really accurate, but it's slightly faster in IE9 mode, and a lil bit faster than that in IE8 mode.

http://jsperf.com/lz-string-performance-enhancement-test

from lz-string.

Thanks for the jsperf. These changes seem to be slightly beneficial to some browser while neutral to others (slightly negative for IE8 but it's IE8 so I quite don't care). In other words it looks good. I'm waiting until tomorrow in order to be able to test this on Windows to make sure I can't find a browser with a negative impact.

from lz-string.

Ah, my bad, I was reading the JSPerf backwards... So it looks like your changes are faster on IE8, neutral on IE>8 and Firefox but slower on Chrome and Safari. Doesn't look good... Unless I'm mistaken once more that is ;-)

from lz-string.

well, the changes were meant to be of benefit to older engines. newer ones are fast enough that it doesn't really matter that much. actually, IIRC, chrome is optimized to handle loops based on string/array length more efficiently. Though, I began wondering after the fact if string length is ever actually recalculated. strings should be immutable, so there'd never be a need to recalculate it like an array does...

from lz-string.

Did you change something on the jsperf? Results are very different this morning... The modified code runs consistently faster than the original one on pretty much all platforms...

from lz-string.

I haven't changed anything to either the jsperf or the library since I posted it. Odd that it should choose to act differently.

from lz-string.

The modified version is 24% slower for me on Chromium 40.

from lz-string.

interesting. I'm not sure how it would've had that much of an effect in any browser, either positive or negative.

from lz-string.

Maybe that's because you are using the master version of lz-string and it got updated? And you haven't rebased your changes.

from lz-string.

Admittedly, I haven't messed with this since the original conversation. It's probably irrelevant at this point.

from lz-string.

Closing this issue. The performance gain wasn't clear or significant and we're going to move toward greater gains with ES6 stuff for newer browsers

from lz-string.